When most people think of "genetic risk," they think of single-gene findings: BRCA1 for breast cancer, APOE4 for Alzheimer's, the APC gene for hereditary colon polyposis. Single high-impact variants exist, and they matter — but they account for only a small fraction of the genetic risk for most common diseases.
The other 90+ percent of genetic risk for things like type 2 diabetes, heart disease, depression, and most cancers is polygenic — distributed across hundreds or thousands of variants, each contributing a tiny amount.
That's where polygenic risk scores come in.
The core idea
A polygenic risk score (PRS) is a single number summarizing your genetic predisposition to a trait or disease, calculated by adding up the effect sizes of hundreds or thousands of variants in your genome.
The math, simplified:
PRS = sum over all variants of (your genotype × effect size from a research study)
Each variant contributes a weighted nudge toward higher or lower risk. Most weights are tiny — single variants typically explain less than 0.1% of disease risk on their own. But added up across thousands of variants, the total can shift your lifetime risk meaningfully.
What "risk" means in a PRS
PRS is usually reported as a percentile in some reference population: "your score is in the 80th percentile for type 2 diabetes" means your genetic risk is higher than 80% of people in that reference population.
Two important nuances:
- Reference matters. A PRS calibrated on European-ancestry data doesn't translate cleanly to other populations. The variants used as inputs are the same, but the effect sizes differ. This is one of the field's biggest open problems.
- Percentile is relative, not absolute. An 80th-percentile PRS for type 2 diabetes maps to a higher absolute lifetime risk than an 80th-percentile PRS for, say, a rare cancer. Always look at both percentile and absolute risk implications.
How a PRS is built
The pipeline:
- Take a large GWAS (genome-wide association study) for the trait — typically 100,000+ participants
- Select SNPs that pass statistical thresholds for association
- Apply quality control, LD pruning (avoiding double-counting variants in linkage disequilibrium)
- Compute weights for each retained SNP from the GWAS
- For a given individual, multiply their genotype at each SNP by its weight and sum
Modern PRS often use 1,000–1,000,000+ SNPs. The "what's the right number" question has been moving — newer methods use far more SNPs with smaller weights, and tend to perform better.
Where PRS works well
The strongest validated PRS in 2026:
- Coronary artery disease — top-decile PRS confers ~3x lifetime risk vs average
- Type 2 diabetes — top-decile PRS ~2-3x vs average
- Atrial fibrillation — ~3x at top decile
- Breast cancer — ~3x at top decile (pre-screening) — meaningfully changes mammography start age
- Inflammatory bowel disease
Where PRS is weaker
PRS for psychiatric conditions, autoimmune diseases, and most cancers has weaker performance — partly because the heritable signal is more dispersed, partly because GWAS sample sizes haven't been as large.
Limitations to know
- Ancestry bias. The vast majority of GWAS data is European-ancestry. PRS performance drops 30-50% in non-European populations. This is a known equity problem the field is actively working on.
- PRS is not deterministic. A high-PRS individual still has below-average outcomes much of the time; lifestyle, environment, and other genetic factors all matter.
- PRS does not replace family history. Family history captures rare high-impact variants that PRS misses; PRS captures common variant load that family history misses. They're complementary.
- Clinical actionability is condition-specific. A high PRS for breast cancer might warrant earlier screening; a high PRS for, say, hair loss has no clinical action.
How Atlagene uses PRS
Atlagene's Premium tier includes polygenic risk scores for the conditions where evidence is strongest. We report:
- Percentile in a reference population
- Absolute lifetime risk implications where evidence supports it
- Ancestry-aware adjustments where validated calibration data exists
- Caveats and limitations in plain language
PRS is one of several signals — single-gene findings, family history, and environmental risk all factor into the dashboard. See methodology for how we grade evidence.
What to do with a high-PRS finding
- Don't panic, don't dismiss. It's a probabilistic signal, not a diagnosis.
- Contextualize with family history. PRS is more meaningful if family history is also positive.
- Discuss with a doctor if the condition has actionable screening (cardiovascular, breast cancer, colon cancer).
- Lifestyle still matters. PRS measures genetic predisposition, not destiny. For most polygenic diseases, lifestyle modification has a similar or larger effect on outcomes than genetics.
Polygenic risk scores are part of Atlagene Premium. See pricing or browse our methodology.